hi there,just to checkwas it so that we use interest rate/convenience yield combo while simulating paths for pricing purposes (i.e. risk-neutral world)? and if we just want to simulate prices for say VaR (or some other use) what drift do we use? or is it just plain martingale?i reckon there was smth on the forum and in my papers but i just can't find it!!!!d.p.s. e.g. would using interest rate differential for drift (in similar fashion to Garman-Kohlhagen) be a plausible idea if we wanted to come up with fx price series? (i bet the question seems silly given that it's almost tantamount to 'how do i forecast FX' ...but what is being used in practice?)

If you are computing VaR for bonds or vanilla interest rate derivatives, you typically don't simulate paths at all. The usual approach is to price each position based on some market factors, and compute VaR using an assumed distribution of those factors.Exotic interest rate derivatives are generally calibrated to vanilla instruments. That implies drift parameters, but they are not set explicitly.If you wanted to compute VaR from scratch for a path-dependent security, you would have to set drift somehow. However, I don't know how you'd get it past a regulator (if that's your goal) because historical data are not useful for estimating drift and deducing it from market prices is highly model-dependent.

thanks aaron!luckily i am not approving anything with regulators, at least in the area of finance it's more like building FX scenarios to show what effects different hedging strategies have on a real bottom-linebut i guess, if i really wanted to use some sort of an interest-rate-differential-based-drift that would be the same as to believe that expected exchange rate is indeed the one predicted with some simple macroeconomic models. and unfortunately in that case i cannot assume that the market did the homework and just calibrate...d.

A standard way of doing var is to simulate paths, but instead of using a process to use historical changes,MJ

For this application, it makes sense to use a simple model of FX rates and calibrate it to vanilla instruments. If you are looking at low-frequency, long-term strategies, setting the drift to the interest rate differential (the simplest consistent model of FX movements) makes sense. If you are looking at high-frequency, short-term strategies, you need to calibrate to swaps and options (and even then the results may not be reliable).A more complex model, one that can price exotic options, will probably give you less reliable results.